The hyperelliptic integrals and \(\pi \) (Q1035599)

The authors consider a class of hyperelliptic integrals involving the square root of \(n\) linear factors of the integration variable. When \(n=3\), the integrals can be evaluated in terms of a Gauss hypergeometric function multiplied by a factor containing \(\pi\); in the cases \(n=4\) and \(n=5\), the integrals are expressed in terms of an Appell function and a Lauricella function, respectively. (These latter functions are higher-order hypergeometric functions involving multiple series.) In the cases considered, the hyperelliptic integrals can also be evaluated alternatively in terms the complete elliptic integrals \({\mathbb K}(k)\) and \({\mathbb E}(k)\) of modulus \(k\). From this two-fold representation of the same mathematical object it is then possible to obtain a representation for \(\pi\) in terms of the above-mentioned special functions. As an example in the case \(n=3\), it is established from the integral \(\int_c^b\sqrt{P(x)}\,dx\), \(P(x)=(b-x)/((a-x)(x-c))\) for \(a>b>c\) that \[ \pi=\frac{4\{(a-c){\mathbb E}(k)+(b-a){\mathbb K}(k)\}}{(b-c)\,{}_2F_1(1/2,1/2;2;k^2)},\qquad k=\left(\frac{b-c}{a-c}\right)^{1/2}. \] Similar representations are obtained for \(n=4\) and \(n=5\) involving the Appell and Lauricella functions in the denominator, respectively. The authors point out that the practical value of such representations does not reside in a computational scheme for the evaluation of \(\pi\), but rather as a benchmark in the computation of the higher-order hypergeometric functions.